Arxiv:1905.04282V2 [Math.NT]

arXiv:1905.04282v2 [math.NT] 5 May 2020 NABDTP OML FOR FORMULA BSD-TYPE A ON References .Aiheial iia wsswt different with twists similar properties Arithmetically basic some of 4. order Summary infinite of points 3.4. Forcing groups Selmer non-trivial 3.3. Forcing 3.2. ..Itrlybetween Interplay applications 3.1. Arithmetic 3. ..Temtv soitdt twist of a Properties to associated motive 2.6. The conjecture period 2.5. Deligne’s 2.4. Periods 2.3. ..Motivic 2.2. curves Motives elliptic of twists 2.1. Artin 2. Notation 1.5. Layout 1.4. ..A rtmtccnetr n t osqecs3 consequences its and conjecture arithmetic 1.3. twists An Artin for formula 1.2. BSD 1.1. Introduction 1. 2010 ahmtc ujc Classification. Subject Mathematics Abstract. lsia ic–wnetnDe conjecture. Birch–Swinnerton-Dyer classical ic–wnetnDe–yefruafor formula Birch–Swinnerton-Dyer–type ieeape f“rtmtclyietcl etnsfre for o associated settings the independently identical” where behave representations, “arithmetically not of do examples group give Tate–Shafarevich the of ersnain.W rnlt xetdpoete of properties expected translate We representations. ehius npriua eehbtstig hr h diff the where settings exhibit we tractable particular be In to appear techniques. not do which group, Tate-Shafarevich rdcin o litccre,adso htteefres force these that show and curves, elliptic for predictions L vle fAtntit felpi curves elliptic of twists Artin of -values LDMRDKHTE,RBR VN,HNEEWIERSEMA HANNEKE EVANS, ROBERT DOKCHITSER, VLADIMIR L -functions hsi nivsiainit h osbeeitneadcon and existence possible the into investigation an is This L ( ,ρ E, p piayprso h aeSaaeihgop15 group Tate-Shafarevich the of parts -primary ) LITCCURVES ELLIPTIC 14 1G5 14G10). (11G05, 11G40 Contents L vle a oehls ie,i otatt the to contrast in differ, nonetheless can -values L fntoso litccre wse yArtin by twisted curves elliptic of -functions 1 L VLE FATNTIT OF TWISTS ARTIN OF -VALUES L vle 23 -values L fntosit ueyarithmetic purely into -functions lpi uvstitdb Artin by twisted curves lliptic m euirpoete fthe of properties peculiar ome erent ytaiinlSle group Selmer traditional by n nte.W also We another. one f p piaycomponents -primary eune fa of sequences 26 22 18 16 15 12 9 9 9 8 7 7 6 5 4 2 2 2 VLADIMIRDOKCHITSER,ROBERTEVANS,HANNEKEWIERSEMA

1. Introduction The Birch–Swinnerton-Dyer conjecture classically provides a connection between the arithmetic of elliptic curves and their L-functions. This link is in many ways still mys- terious. Indeed, some properties of L-functions do not obviously correspond to arithmetic properties of elliptic curves and vice versa, a classical example being the compatibility of the conjecture with isogenies, which is a highly non-trivial theorem of Cassels. In this article we focus on factorisation of L-functions: when E/Q is an elliptic curve and F/Q a finite extension, L(E/F,s) factorises as a product of L-functions of twists of E by Artin representations L(E,ρ,s). We investigate what standard conjectures say specifically for these twisted L-functions. Ideally, we would like to give a BSD-type formula for the leading term at s=1 for L(E,ρ,s), but, as we shall explain, there is a significant barrier to this. However, we shall provide a tool for extracting explicit arithmetic predictions, and illustrate its use by exhibiting new phenomena about the behaviour of Tate–Shafarevich groups, Selmer groups and rational points. 1.1. BSD formula for Artin twists. The Birch–Swinnerton-Dyer conjecture states that

ords=1 L(E/F,s)=rk E/F, and that the leading term of the Taylor series at s=1 of the L-function is given by X L(E/F,s) ∆ Reg E/F CE/F F E/F | | ( ) lim r r +|r | r = 2 , † s 1 (s 1) · Ω+(E) 1 2 Ω (E) 2 E(F )tors → − p | − | | | where r is the order of the zero, (r1, r2) is the signature of F , Ω are the periods of E ± and CE/F is the product of Tamagawa numbers and other local fudge factors from finite places (see 1.5). Of course, the formula implicitly assumes that the Tate–Shafarevich group X § E/F is finite. We will refer to the expression on the right-hand side of ( ) as BSD(E/F ). Just as the Dedekind ζ-function can be expressed as a product of Artin L†-functions, so the L-function L(E/F,s) can be written as a product of twisted L-functions L(E,ρ,s) for Artin representations that factor through the Galois closure of F/Q. The (conjectural) analogue of the Birch–Swinnerton-Dyer rank formula is well known in this context (see e.g. [10] 2): § Conjecture 1. For an elliptic curve E/Q and an Artin representation ρ over Q, ord L(E,ρ,s)= ρ, E(K) . s=1 h Ci Here, and throughout, E(K)C = E(K) Z C where K is any finite Galois extension of Q such that ρ factors through Gal(K/Q), and ⊗, denotes the usual representation theoretic inner product of characters. In other words,h· the·i conjecture predicts that, for an (irreducible) ρ, the order of vanishing of L(E,ρ,s) is the “multiplicity” of ρ in the group of K-rational points of E. However, the situation with the second part of the Birch–Swinnerton-Dyer conjecture appears to be much more difficult. Problem 2. Formulate a BSD-like formula for the leading term at s=1 of L(E,ρ,s). There appears to be a barrier to finding such an expression, as there are “arithmetically identical” settings giving rise to different L-values. We write L (E,ρ) for the modification of the leading term of L(E,ρ,s) analogous to the left-hand side of ( ) (see Definition 12). † Example 3 (see also 4). The elliptic curves with Cremona labels E =307a1 and E′ =307c1 have the same conductor,§ same discriminant, no rational points, trivial Tate–Shafarevich + group and trivial local Tamagawa numbers both over Q and over Q(ζ11) . However, for a Dirichlet character χ of order 5 and conductor 11, L (E,χ) = L (E′,χ). Specifically, 1 √5 2 6 L (E,χ) = 1, while L (E′,χ) = ( ± ) , the sign of √5 depending on the choice of χ. 2 ± ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 3

1.2. An arithmetic conjecture and its consequences. We will not propose an exact expression for the hypothetical BSD(E,ρ) term for the conjectural formula L (E,ρ) = BSD(E,ρ). However, based on the behaviour of L-functions, we will show that BSD(E,ρ) must satisfy the list of properties given in Conjecture 4 below. One of the roles of BSD(E,ρ) is that it lets one decompose the Birch–Swinnerton-Dyer quotient BSD(E/F ) according to Artin representations, analogously to the factorisation of L-functions. This may at first glance look almost vacuous, but, as we will explain, the existence of such a decomposition has a range of consequences for Selmer groups, Tate–Shafarevich groups and ranks of elliptic curves. We write Q(ρ) for the field generated by the values of the character of ρ, write ρ∗ for the dual representation, and wρ and wE,ρ for the root number of ρ and of the twist of E by ρ, respectively. Conjecture 4. Let E/Q be an elliptic curve. For every Artin representation ρ over Q there is an invariant BSD(E,ρ) C× with the following properties. Let ρ and τ be Artin representations that factor through∈ Gal(K/Q). X C1. BSD(E/F ) = BSD(E, IndF/Q 1) for a number field F (and E/F is finite). C2. BSD(E,ρ τ) = BSD(E,ρ) BSD(E, τ). ⊕ r 2 C3. BSD(E,ρ) = BSD(E,ρ∗) ( 1) w w− , where r = ρ, E(K) . · − E,ρ ρ h Ci C4. If ρ is self-dual, then BSD(E,ρ) R and sign BSD(E,ρ) = sign w . ∈ ρ If ρ, E(K)C =0, then moreover: h i g g C5. BSD(E,ρ) Q(ρ)× and BSD(E,ρ ) = BSD(E,ρ) for all g Gal(Q(ρ)/Q). C6. If ρ is a non-trivial∈ primitive Dirichlet character of order d, ∈and either the conductors of E and ρ are coprime or E is semistable and has no non-trivial isogenies over Q, then BSD(E,ρ) Z[ζ ]. ∈ d Theorem 5 (see Corollary 25). Conjecture 4 holds assuming the analytic continuation of L-functions L(E,ρ,s), their functional equation, the Birch–Swinnerton-Dyer conjecture, Deligne’s period conjecture, Stevens’s Manin constant conjecture for E/Q and the Riemann hypothesis for L(E,ρ,s). Note that the statement of the conjecture is free of L-functions. Morally, it should be purely a property of Selmer groups. However, it has some consequences that do not appear to be tractable with classical Selmer group techniques, as we now illustrate. Theorem 6 (see Theorem 28, Example 29). Let ℓ and p be primes such that the primes above p in Q(ζℓ) are non-principal and have residue degree 2. If Conjecture 4 holds then, for X every semistable elliptic curve E/Q with no non-trivial isogenies, E/Q[p] =1 and cv =1 for all rational primes v, and for every cyclic extension F/Q of degree| ℓ with| E(F )= E(Q), 2 if X [p∞] = p then X [q∞] =1 for some q = p. | E/F | | E/F | 6 6 Roughly speaking, in the setting of the theorem the presence of the p-primary part of X forces some other part of X to be non-trivial too. It would be interesting to have a purely Selmer theoretic method that can explain such behaviour. Conjecture 4 can also be used to show that purely local constraints can force certain Selmer groups of E over extensions F/Q to become non-trivial. More usual methods for achieving such criteria either use Galois module structures or Iwasawa theoretic methods (both can be used to make Selp(E/F ) non-trivial for p [F : Q], see e.g. [1, 9]) or use some form of the parity conjecture (this requires [F :Q] to be| even). 4 VLADIMIRDOKCHITSER,ROBERTEVANS,HANNEKEWIERSEMA

Theorem 7 (see Corollary 31). Suppose Conjecture 4 holds. There is an (explicit) Galois number ﬁeld F of odd degree and (explicit) rational prime ℓ, such that every elliptic curve E/Q with additive reduction at ℓ of Kodaira type III and good reduction at other primes that ramify in F/Q has a non-trivial p-Selmer group Selp(E/F ) for some prime p ∤ [F :Q]. Finally, we will also show that Conjecture 4 can be used to establish purely theoretical results, such as the following case of the Birch–Swinnerton-Dyer conjecture for twists of elliptic curves by dihedral Artin representations (below D2pq denotes the dihedral group of order 2pq). As far as we are aware, this does not follow from known cases of the parity conjecture.

Theorem 8 (see Theorem 35). Let F/Q be a Galois extension with Galois group D2pq, with p, q 3 mod 4 primes, and let ρ be a faithful irreducible Artin representation that factors through≡ F/Q. If Conjecture 4 holds, then for every semistable elliptic curve E/Q, if ord L(E,ρ,s) is odd, then ρ, E(F ) > 0. s=1 h Ci We stress once again that Conjecture 4 ought to be purely a statement about Selmer groups, although we do not understand the extra structure on Selmer groups or on X that causes it: our justification of the conjecture relies on L-functions. For applications like Theorem 8 it is clearly important to find a proof that does not assume the Birch–Swinnerton- Dyer conjecture. Problem 9. Justify Conjecture 4 without assuming the Birch–Swinnerton-Dyer conjecture. Remark 10. The conjecture completely determines the value of BSD(E,ρ) for Artin representations ρ whose character is Q-valued. Indeed, for a finite group G, the image of the Burnside ring in the rational representation ring has finite index. Thus, if ρ factors through Gal(K/Q) where K/Q is a finite Galois extension, there are intermediate fields Fi, Fj′ of m K/Q and a positive integer m such that ρ⊕ IndF /Q 1 IndF ′/Q 1, and so (C1), ⊕ i i ≃ j j (C2) and (C4) imply that BSD(E,ρ) is the unique real number such that L L m j BSD(E/Fj′) BSD(E,ρ) = and signBSD(E,ρ) = sign wρ. Q i BSD(E/Fi) Remark 11. As illustrated inQ Theorem 8 above, our method sometimes allows us to predict the existence of points of infinite order on elliptic curves (see also 3.3, and Theorem 33 for an example with a quaternion Galois group). Needless to say, we§ have not found a setting where we can predict the existence of rational points which is not already predicted by the parity conjecture, or which outright contradicts it. The computations arising in our approach look very different from the theory of local root numbers, but (rather magically) always match. 1.3. L-values of Artin twists of elliptic curves. The heart of our approach to deriving Conjecture 4 lies in extracting precise consequences of L-function conjectures in the setting of Artin twists of elliptic curves. For the “L-function side” of the sought Birch–Swinnerton- Dyer formula for twists we use the following modification of the leading term of L(E,ρ,s) at s = 1. This is very carefully chosen so as to mesh well with the Birch–Swinnerton-Dyer conjecture over number fields, the functional equation and Deligne’s period conjecture for Artin twists of elliptic curves (see 2.4) at the same time. We will show that it satisfies the analogues of (C1)–(C6) of Conjecture§ 4, which is our justification for the conjecture. Definition 12. For an elliptic curve E/Q and an Artin representation ρ over Q, we write f L L(E,ρ,s) ρ (E,ρ) = lim r d+(ρ) d−(ρ) , s 1 (s 1) · Ω+(E) Ω (E) wρ → − p| − | ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 5 where r = ords=1 L(E,ρ,s) is the order of the zero at s = 1, fρ is the conductor of ρ, and d±(ρ) are the dimensions of the 1-eigenspaces of complex conjugation in its action on ρ. ± Theorem 13 (Theorem 24, Corollary 26). Let E/Q be an elliptic curve and let ρ be an 2 2 1 Artin representation over Q. Fix ζ satisfying ζ = wρwE,ρ− . Suppose that for all Artin representations ψ over Q, the L-functions L(E,ψ,s) have analytic continuation to C and satisfy the functional equation, Deligne’s period conjecture and the Riemann hypothesis. Suppose also that Stevens’s Manin constant conjecture holds for E/Q and the Birch–Swinnerton-Dyer conjecture holds for E over number fields. Then

(1) L (E, IndF/Q 1) = BSD(E/F ) for a number field F . (2) L (E,ρ ρ′)= L (E,ρ)L (E,ρ′). ⊕ r 2 (3) L (E,ρ∗) = ( 1) ζ L (E,ρ), where r = ord L(E,ρ,s). − s=1 (4) If ρ ρ∗ then L (E,ρ) R and w L (E,ρ) > 0. ≃ ∈ ρ · Henceforth suppose that moreover L(E, ρ, 1) =0. Then 6 (5) L (E,ρ) Q(ρ). ∈ (6) L (E,ρ) Q(ρ) is invariant under complex conjugation as a fractional ideal of Q(ρ). (7) L (E,ρg)=· O L (E,ρ)g for all g Gal(Q(ρ)/Q). − ∈ 2 d (ρ) dim ρ (8) ζ is a root of unity. If fE is coprime to fρ, then ζ = ( 1) wE det ρ(fE), where det ρ is regarded as a primitive Dirichlet character (see− Notation 15). + 1 (9) ζ L (E,ρ) Q(ρ, ζ) ; in particular arg L (E,ρ) = arg ζ− . · ∈ ± (10) If ρ is a non-trivial primitive Dirichlet character of order d, and either fρ is coprime to f or E is semistable and has no non-trivial isogenies over Q, then L (E,ρ) Z[ζ ]. E ∈ d ′ m Qj BSD(E/Fj ) 1 Let B = for any number fields Fi, Fj′ and positive integer m that satisfy Qi BSD(E/Fi) g rm ′ ( g G ρ ) i IndFi/Q 1 = j IndF /Q 1, where G = Gal(Q(ρ)/Q). Then ∈ ⊕ j (11)L N (LL(E,ρ)) = B,L with sign + if m is odd. Q(ρ)/Q ± (12) N + (ζ L (E,ρ)) = √B if ρ ρ∗ and ζ Q(ρ). Q(ρ) /Q · ± 6≃ ∈ (13) N + (ζ L (E,ρ)) = B if ρ ρ∗ and ζ / Q(ρ). Q(ρ,ζ) /Q · ± 6≃ ∈ Remark 14. The main reason why most of the results above require the assumption that the L-value is non-zero is that it features in Deligne’s period conjecture. It might be possible to extend the predictions to the higher rank case using the motivic L-value conjectures (Beilinson, Bloch–Kato, Equivariant Tamagawa Number Conjecture). These may also let one generalise the integrality statement (10) to other Artin representations and to pin down the ideal generated by L (E,ρ) more precisely. We will not attempt to address this here. 1.4. Layout. This paper is split into three parts. In 2 we extract the explicit L-value predictions of Theorem 13 from the classical conjectures§ and deduce Theorem 5 from them. The key technical step here is to express the periods associated to an Artin twist of an elliptic curve to the classical periods Ω (Corollary 23). In 3 we develop the arithmetic consequences for elliptic curves, includin± g Theorems 6, 7 and 8.§ The main ingredient is Theorem 27, which, based on Conjecture 4, lets us link easily controllable local invariants to ranks and the Tate–Shafarevich group. In view of Theorem 5 these results are all consequences of the classical conjectures on L-functions. In 4 we discuss explicit examples of L-values of twists of elliptic curves by Dirichlet characters§ and illustrate the difficulty of refining Theorem 13 to a clean BSD-type prediction for the value of L (E,ρ). We end by giving several tables of examples of a similar kind to Example 3.

1These exist by Remark 10. 6 VLADIMIRDOKCHITSER,ROBERTEVANS,HANNEKEWIERSEMA

We have kept the three sections largely independent of one another. In particular, the reader who does not wish to grapple with the motivic background can skip directly to the arithmetic applications in 3 or the L-value examples in 4. § § 1.5. Notation. We fix (once and for all) an algebraic closure Q inside C. All our number fields will be subfields of this choice of Q. Formally, all our Artin representations will be C-valued; that is, defined by a group homomorphism ρ : GK AutC(V ) that factors through Gal(F/K) for some finite Galois extension F/K and some→ finite dimensional complex vector space V . We will typically work with isomorphism classes of Artin representations, without explicitly mentioning it. The following notation is used throughout the paper: E an elliptic curve defined over Q. cv(E/F ) the local Tamagawa number of E/Fv. G the absolute Galois group Gal(Q/F ) of a number field F Q. F ⊆ Frobp (arithmetic) Frobenius element at a prime p. ρ∗ the dual representation of an Artin representation ρ. Q(ρ) the (abelian) extension of Q generated by the character values of ρ. ρg for g Gal(Q(ρ)/Q), the Artin representation with character Tr ρg = g Tr ρ. ∈ ◦ d±(ρ) the dimension of the 1-eigenspace of complex conjugation on ρ. ± fρ the Artin conductor of ρ. fE the conductor of E/Q. wρ the Artin root number of ρ. wE the root number of E/Q (the sign in the functional equation). w the root number of the twist E/Q by ρ (see 2.5). E,ρ § Ind ρ IndGK ρ for a field extension F/K and an Artin representation ρ over F. F/K GF the formal difference of Artin representaions, i.e. ρ ρ = ρ ρ = ρ ρ . ⊖ 1 ⊖ 2 3 ⇔ 1 2 ⊕ 3 ζn a primitive n-th root of unity. NF/K the norm map from F to K. Notation 15. We use the convention (as in [5] 3.2) that the Euler factor at a prime p of L(E,ρ,s) is § 1 s 1 I det Id Frob− p− (H (E) ρ) p , − p ℓ ⊗ where I is the inertia group at p, H1(E)= H1 (E, Q ) C for any embedding Q ֒ C p ℓ et ℓ Qℓ ℓ and any prime ℓ = p. ⊗ → To identify 1-dimensional6 Artin representations with Dirichlet characters, we use the a isomorphism Gal(Q(ζn)/Q) (Z/nZ)× given by σa a for σa : ζn ζn. → ↔ → s We caution the reader that with these normalisations, if L(E,s) = ann− and χ is a primitive Dirichlet character of conductor coprime to that of E, then P ∞ s L(E,χ,s)= χ(n)ann− . n=1 X Notation 16. For an elliptic curve E/Q, we define the -periods of E to be ±

Ω+(E)= ω and Ω (E)= ω, + − − ZE(C) ZE(C) where ω is a global minimal diﬀerential on E and E(C)± is the set of points P E(C) such ∈ that P¯ = P with orientation chosen so that Ω+(E) R>0 and Ω (E) iR>0. ± ∈ − ∈ ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 7

Notation 17. For an elliptic curve E/Q and a number field F, we define ω C = c (E/F ) , E/F v ωmin v v v Y where v runs over the finite places of F, ω is a global minimal differential for E/Q and min ω ωv is a minimal differential at v. By min we mean any scalar λ F × that satisfies ωv min ∈ ω = λωv . In terms of minimal discriminants, if E is given by a Weierstrass equation 2 3 2 dx y + a1xy + a3 = x + a2x + a4x + a6 with discriminant ∆E and ω = , then 2y+a1x+a3 12 ω − ∆E min = min . ωv v ∆ v E,v

Notation 18. For an elliptic curve E/Q and a number field F, we define X Reg E/F CE/F BSD(E/F )= E/F | | . E(F ) 2 | tors| We also briefly recall Stevens’s version of the Manin constant conjecture ([12] Conj. I): Conjecture 19 (Stevens’s Manin constant conjecture). Every elliptic curve over Q of conductor N admits a modular parametrisation X (N) E with Manin constant 1. 1 → Acknowledgements. The authors would like to thank Chris Wuthrich for pointing out an issue in our original claim about integrality of L-values and for fixing it in [16], and David Burns for his valuable comments on a draft of the present article. The first named author was supported by a Royal Society University Research Fellowship. The third named author was supported by the Engineering and Physical Sciences Research Council [EP/L015234/1], through the EPSRC Centre for Doctoral Training in Geometry and Number Theory (the London School of Geometry and Number Theory) at University College London.

2. Artin twists of elliptic curves In order to explain the implications of Deligne’s period conjecture for Artin twists of elliptic curves, we ﬁrst recall the relevant deﬁnitions from the theory of motives. We shall follow closely the presentations given in [4, 4] and [15, 2] and refer the reader to Deligne’s article [7] for a more detailed account. § § Notation 20. The following additional notation applies only in this section:

ι the element of GQ corresponding to complex conjugation. F a number field (inside our fixed algebraic closure Q, as always). F the ring F C (unadorned tensor products are over Q). C ⊗ Fℓ the ring F Qℓ λ ℓ Fλ where Fλ is the completion of F at the prime λ. ⊗ ≃ | ΣF the set of real and complex embeddings F C. Q → ε(ρ) the epsilon factor of an Artin representation ρ at s = 0, i.e. ε(ρ)= wρ fρ. 2.1. Motives. It will be sufficient for our purposes to view motives in the naivep sense; that is, as a collection of vector spaces with certain additional structures and comparison isomorphisms between them. In particular, a (homogeneous) motive M over Q with coefficients in a number field F, dimension d and weight w carries the following data:

(1) a) A d-dimensional F-vector space HB(M) (the Betti realisation). b) An F-linear involution F on HB(M). ∞ 8 VLADIMIRDOKCHITSER,ROBERTEVANS,HANNEKEWIERSEMA

c) A Hodge decomposition into free FC-modules: H (M) C = Hr,s(M) B ⊗ r+s=w M such that F Hr,s(M)= Hs,r(M). ∞

(2) a) A d-dimensional F-vector space HdR(M) (the de Rham realisation). b) A decreasing ﬁltration F kH (M): k Z of F-subspaces of H (M). { dR ∈ } dR

(3) a) For each prime ℓ, a free Fℓ-module Hℓ(M) of rank d (the ℓ-adic realisation). b) For each prime ℓ, a continuous action of GQ on Hℓ(M).

(4) A comparison isomorphism between FC-modules

IM, : HB (M) C ∼ HdR(M) C ∞ ⊗ −→ ⊗ such that IM, (F ι) = (id ι) IM, and ∞ ◦ ∞ ⊗ ⊗ ◦ ∞ r,s k IM, H (M) = F HdR(M) C. ∞ ⊗ r k M≥ Remark 21. Comparison isomorphisms between other realisations are also part of the data carried by M; however, as we shall not need these for the work that follows, we choose to omit them here and refer the interested reader to sections 2.5 and 2.6 of [15]. 2.2. Motivic L-functions. Let M be a motive over Q with coeﬃcients in F. For any prime number ℓ, identifying Fℓ with λ ℓ Fλ gives rise to a decomposition |

Q Hℓ(M)= Hλ(M), λ ℓ M| where Hλ(M) is the image of Hℓ(M) under scalar multiplication by unity in Fλ. For each prime number p, let Dp GQ be a choice of decomposition group at p and let I D be the corresponding inertia⊆ subgroup. The local polynomial of M at p is p ⊆ p 1 I P (M,t) = det id t Frob− H (M) p , p − p | λ where λ is a prime of F not lying over p. We assume the standard hypothesis that Pp(M,t) is independent of the choice of λ and has coefficients in F. For each σ ΣF, we define ∈ s 1 L(σ,M,s)= σP (M,p− )− C, p ∈ p Y where σPp(M,X) σF[X] C[X]; the expression converges for s with sufficiently large real part. It is conjectured∈ that⊂ each L(σ,M,s) admits a meromorphic continuation to the entire complex plane which satisfies a functional equation of the form

L (M,s)L(σ,M,s)= ε(σ,M,s)L (M ∗, 1 s)L(σ, M ∗, 1 s), ∞ ∞ − − where the Euler factor at inﬁnity L (M,s) is a product of gamma functions which does not depend on σ (see [7, Proposition∞ 2.5]) and the epsilon factor ε(σ,M,s) is a product of a constant and an exponential (see [14] for the details and extra hypotheses required for this construction). ΣF It is convenient, by identifying FC with C via the canonical isomorphism of C-algebras:

x z (zσ(x): σ ΣF), ⊗ 7−→ ∈ to form a single L-function associated with M which takes values in FC :

L(M,s) = (L(σ,M,s): σ ΣF). ∈ ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 9

2.3. Periods. Let M be a motive over Q with coeﬃcients in F. For simplicity, we shall restrict to the case where M has odd weight w. Let HB(M)± denote the 1-eigenspaces of the endomorphism F and let ± ∞ HdR(M) HdR(M)± = 1+ w/2 F ⌊ ⌋HdR(M) for both choices of sign. The -period map α± of M is the composition of the following ± M FC-linear maps:

H (M)± C H (M) C ∼ H (M) C H (M)± C, B ⊗ −→ B ⊗ −→ dR ⊗ −→ dR ⊗ where the ﬁrst map is induced by inclusion, the second map is the Betti-de Rham comparison isomorphism, and the last map is induced by the natural quotient map. It follows from

[7, 1.7] that αM± is an isomorphism. The -period of M, denoted by c±(M), is deﬁned to be the§ residue class ±

det(αM± ) mod F× in FC×/F×, where the determinant of the -period map is calculated with respect to F-bases. ΣF ± As above, by identifying FC with C , we can also view c±(M) as a ‘tuple’:

c±(M) = (c±(σ, M) C×/F× : σ ΣF). ∈ ∈ 2.4. Deligne’s period conjecture. Let M be a motive over Q with coefficients in F. We retain the assumption that M has odd weight. We say that M is critical (at s =0) if, whenever j < k and Hj,k(M) = 0, one has j < 0 and k 0. See [4, Lemma 3] for a proof that this is equivalent to the definition6 of criticality given≥ in [7]. + Suppose that M is critical and fix a choice of representative for the period c (M) in FC×. Then conjectures 2.7 and 2.8 of [7] assert that

(1) ord L(σ,M,s) is independent of σ ΣF and is non-negative. s=0 ∈ (2) If L(M, 0) =0, there exists x F× such that, for all σ ΣF, one has 6 ∈ ∈ L(σ, M, 0) = σ(x)c+(σ, M).

2.5. The motive associated to a twist. Let E be an elliptic curve over Q and let ρ be an Artin representation over Q. Choose any finite abelian extension F/Q over which ρ can be realised and let τ be an F-linear representation of GQ such that C F τ ρ. In order to understand Deligne’s period conjecture in the setting of Artin⊗ ≃ twists of elliptic curves, we are led to consider the tensor product motive h1(E)(1) [τ], whose associated realisations and comparison isomorphisms arise by taking the tensor⊗ product of the corresponding data for the motives h1(E)(1) and [τ] (see Examples 2.1B and 2.1C in [15] for detailed information about the latter two motives). In particular, one has that L(h1(E)(1) [τ],s) = (L(E,ργ,s + 1) : γ Gal(F/Q)), ⊗ ∈ where L(E,ρ,s) is the Artin-twisted Hasse-Weil L-function whose construction is described explicitly in [5, 3.2]. We recall (loc. cit.) that L(E,ρ,s) is conjectured to admit an analytic continuation to§ the whole complex plane which satisfies a functional equation of the form dim ρ dim ρ Γ (s) L(E,ρ,s)= ε(E,ρ,s)Γ (2 s) L(E,ρ∗, 2 s), C C − − s 1 s where Γ (s) = 2(2π)− Γ(s) and the epsilon factor ε(E,ρ,s) has the form w N − where C E,ρ · E,ρ wE,ρ C has absolute value 1 (the root number of the twist) and NE,ρ is a positive integer (the conductor∈ of the twist). Finally, we recall that one has the following “Artin formalism”: (1) L(E,ρ ρ ,s)= L(E,ρ ,s)L(E,ρ ,s), 1 ⊕ 2 1 2 (2) L(E, IndF/Q 1,s)= L(E/F,s), 10 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA where L(E/F,s) is the usual (‘un-twisted’) Hasse-Weil L-function of E/F. The following theorem will allow us to find an explicit representative for the period c+(h1(E)(1) [τ]) in terms of the periods associated with the motives h1(E)(1) and [τ]. ⊗ Theorem 22. Let M be a motive over Q with rational coefficients such that (1) M has dimension 2 and weight 1, − (2) dimQ HB(M)± =1, 0, 1 1,0 (3) H (M) C = H − (M) H− (M). B ⊗Q ⊕ Let N be a motive over Q with coefficients in a number field F such that (1) N has dimension d and weight 0, (2) H (N) C = H0,0(N). B ⊗Q Under these conditions, the motive M N is critical and ⊗ + + µ ν c (M N)= c (M) c−(M) det(IN, ) mod F×, ⊗ ∞ + where µ = dimF(HB(N) ), ν = dimF(HB(N)−) and det(IN, ) is computed using F-bases. ∞ Proof. The tensor product motive M N is specified by the data obtained by taking the tensor product of the realisations of M⊗and N and their additional structures; in particular, M N is a motive of dimension 2d and weight 1 such that ⊗ − (a) H (M N)= H (M) H (N) as an F-vector space. B ⊗ B ⊗Q B (b) F (M N)= F (M) Q F (N) as an F-linear involution. ∞ ⊗ ∞ ⊗ ∞ (c) HdR(M N)= HdR(M) Q HdR(N) as an F-vector space. (d) The de Rham⊗ filtration on⊗H (M N) is dR ⊗ HdR(M N) if k 1, F kH (M N)= F 0H (M⊗ ) H (N) if k =0≤−, dR  dR Q dR ⊗ 0⊗ if k 1.  ≥ (e) The Betti-de Rham comparison isomorphism IM N, is  ⊗ ∞ I ∞ CI ∞ H (M) H (N) C M, ⊗ N, H (M) H (N) C B ⊗Q B ⊗Q −−−−−−−−−→ dR ⊗Q dR ⊗Q viewed as an isomorphism of FC-modules, where we have identified H (M) H (N) C = H (M) C H (N) C B ⊗Q B ⊗Q B ⊗Q ⊗C B ⊗Q and similarly for the de Rham realisations. It follows easily from properties (a)-(d) that H (M) H (N) H (M N)+ = dR ⊗Q dR dR ⊗ F 0H (M) H (N) dR ⊗Q dR HdR(M) = 0 Q HdR(N) F HdR(M) ⊗ = H (M)+ H (N), dR ⊗Q dR + H (M N)+ = H (M) H (N) B ⊗ B ⊗Q B + + = H (M) H (N) H (M)− H (N)− . B ⊗Q B ⊕ B ⊗Q B We choose bases for our various spaces as follows: + + (1) a Q-basis γ (resp. γ− ) for HB(M) (resp. HB(M)−), { } 0 { } (2) a Q-basis ω0 for F HdR(M) and extend to a basis ω0,ω1 for HdR(M), { +} + { + } (3) an F-basis v ,...,v (resp. v−,...,v− ) for H (N) (resp. H (N)−), { 1 µ } { 1 ν } B B ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 11

(4) an F-basis w ,...,w for H (N). { 1 d} dR In terms of these bases, we have

IM, (γ± 1) = ω0 η0± + ω1 η1± for some η0±, η1± C, ∞ ⊗ ⊗ ⊗ ∈ d

IN, (vj± 1) = (bij± ξij±)(wj 1) for some bij± ξij± FC, ∞ ⊗ ⊗ ⊗ ⊗ ∈ i=1 X and so it follows from (e) that

d d

IM N, (γ± v± 1) = (b± η0±ξ±)(ω0 wi 1)+ (b± η1±ξ±)(ω1 wi 1). ⊗ ∞ ⊗ j ⊗ ij ⊗ ij ⊗ ⊗ ij ⊗ ij ⊗ ⊗ i=1 i=1 X X + Hence, with respect to these bases, the matrix of αM N has ijth component: ⊗ + + (1 η1 )(bij ξij ) if j µ, Aij = ⊗ ⊗ + ≤ (1 η−)(b ξ ) if µ < j n, ⊗ 1 ij ⊗ ij ≤ and so taking determinant yields the desired expression:

+ + µ ν c (M N)=(1 η1 ) (1 η1−) det(IN, ) mod F× ⊗ ⊗ ⊗ ∞ + µ ν = c (M) c−(M) det(IN, ) mod F×. ∞ Finally, to see that M N is critical, we simply observe that ⊗ 0, 1 1,0 H (M N) C = H − (M N) H− (M N), B ⊗ ⊗ ⊗ ⊕ ⊗ where, viewed as FC-modules, we have 0, 1 0, 1 0,0 H − (M N)= H − (M) H (N), ⊗ ⊗C 1,0 1,0 0,0 H− (M N)= H− (M) H (N). ⊗ ⊗C Corollary 23. Let E be an elliptic curve over Q and ρ be an Artin representation over Q. Let F/Q be a ﬁnite abelian extension over which ρ can be realised and let τ be an F-linear 1 representation of G such that C F τ ρ. Then h (E)(1) [τ] is a critical motive and the Q ⊗ ≃ ⊗ component of c+ h1(E)(1) [τ] corresponding to our ﬁxed embedding F Q C is ⊗ ⊆ ⊆ − d+(ρ) d (ρ) Ω+(E) Ω (E) wρ | − | mod F×. fρ Proof. Applying Theorem 22 with M =ph1(E)(1) and N = [τ] yields

− + 1 d+(τ) d (τ) c h (E)(1) [τ] =Ω+(E) Ω (E) det(Iτ, ) mod F×. ⊗ − ∞ Moreover, it follows from [7, 5.6.1] that

− d (τ) γ i det(Iτ, ) = (ε(τ ): γ Gal(F/Q)) mod F×, ∞ ∈ − − − d (τ) d (τ) d (τ) and so, since d±(τ)= d±(ρ),ε(τ)= ε(ρ) and Ω (E) = i Ω (E) , we see that − | − | − d+(ρ) d (ρ) Ω+(E) Ω (E) ε(ρ) mod F× | − | is equal to the component of c+(h1(E)(1) [τ]) corresponding to the identity in Gal(F/Q). ⊗ Therefore, since ε(ρ)= wρ fρ, the result follows on dividing through by fρ. p 12 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA

2.6. Properties of L (E,ρ). We now turn to L-values and the proof of Theorem 13.

Theorem 24. Let E/Q be an elliptic curve and let ρ and ρ′ be Artin representations over Q. Suppose that L(E,ρ,s) and L(E,ρ′,s) admit an analytic continuation to C.

L1. If ρ = IndF/Q 1 for a number ﬁeld F , then

L L(E/F,s) ∆F (E,ρ) = lim r r +|r | r , s 1 (s 1) · Ω+(E) 1 2 Ω (E) 2 → − p | − | where (r1, r2) is the signature of F and r = ords=1 L(E/F,s). L2. L (E,ρ ρ′)= L (E,ρ)L (E,ρ′). L3. If the functional⊕ equation for L(E,ρ,s) holds near s =1, then r L ( 1) wE,ρ L (E,ρ)= − 2 (E,ρ∗), wρ

where r = ords=1 L(E,ρ,s). L4. If ρ is self-dual i.e. ρ ρ∗, then ≃ (i) L (E,ρ) R, ∈ (ii) sign L (E,ρ) = sign wρ, providing the Riemann Hypothesis holds for L(E,ρ,s). L5. If L(E, ρ, 1)=0 and Deligne’s period conjecture holds for the twist of E by ρ, then 6 (i) L (E,ρ) Q(ρ)×, (ii) L (E,ρg)=∈ L (E,ρ)g for all g Gal(Q(ρ)/Q). ∈ L6. If ρ is a non-trivial primitive Dirichlet character of order d, and either (i) E is semistable and has no non-trivial isogenies over Q, or (ii) Stevens’s Manin constant conjecture holds for E/Q and fρ is coprime to fE, then L (E,ρ) Z[ζ ]. ∈ d Proof. For any Artin representation ρ over Q, we shall denote the leading coefficient in the Taylor series expansion of L(E,ρ,s) at s = 1 by LT(E, ρ, 1). In this notation, Definition 12 states that f LT(E, ρ, 1) L ρ (E,ρ)= d+(ρ) d−(ρ) . Ω+(E) Ω (E) wρ p | − | Recall (from [8], for example) that for an Artin representation ρ over a number field F , the conductor fρ and root number wρ are, respectively, an integral ideal of F and a complex number of absolute value 1, and that they have the following formal properties:

(1) fρ ρ = fρ fρ and wρ ρ = wρ wρ , 1⊕ 2 1 2 1⊕ 2 1 2 dim ρ (2) f IndL/F (ρ) = disc(L/F ) NL/F (fρ) and w IndL/F (ρ) = wρ. We refer to these as “Artin formalism”, analogously to the case of L-functions given in 2.5. § L1. By Artin formalism for the other factors, it suﬃces to prove that + d (IndF/Q 1)= r1 + r2 and d−(IndF/Q 1)= r2.

Since IndF/Q 1 is the permutation module GQ/GF , we have + + d + d− = [F : Q] and d d− = tr Ind (1)(ι) , − F/Q where ι GQ is complex conjugation. However, we also have that ∈ tr IndF/Q(1)(ι) = # singleton orbits of ι in GQ/GF = r1, + + and so d + d− =r1 +2r2 and d d− = r1 and the claim now follows. − ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 13

L2. By Artin formalism for the other factors, it suﬃces to note the identity:

d±(ρ ρ′)= d±(ρ)+ d±(ρ′). ⊕ dr L3. Applying to the functional equation for L(E,ρ,s) yields dsr s=1 r LT(E, ρ, 1) = wE,ρ ( 1) LT(E,ρ∗, 1), · − and so, since fρ∗ = fρ and d±(ρ∗)= d±(ρ), we have

∗ ∗ fρ LT(E, ρ, 1) r wE,ρwρ fρ LT(E,ρ∗, 1) d+(ρ) d−(ρ) = ( 1) d+(ρ∗) d−(ρ∗) Ω+(E) Ω (E) wρ − wρ Ω+(E) Ω (E) wρ∗ p | − | p | − | which, on recalling that wρwρ∗ = wρ ρ∗ =1, simplifies to the given formula. ⊕ L4. For s C with Re(s) 1,L(E,ρ,s) can be expressed in terms of its Dirichlet series and since the character∈ of ρ is≫ real-valued it follows that the coefficients of this series are real. In particular, one has L(E,ρ,s)= L(E, ρ, s) for all sufficiently large s R. Since L(E,ρ,s) is ∈ analytic everywhere in C (conjecturally), it follows that L(E,ρ,s)= L(E, ρ, s) for all s C ∈ and so, in particular, that LT(E, ρ, 1) R. Moreover, since ρ is self-dual, we have wρ = 1 and so it follows that L (E,ρ) R. ∈ ± Assuming the Riemann Hypothesis∈ holds, we have that L(E,ρ,s) = 0 for all s (1, ). Moreover, by looking at the Euler product, it is clear that L(E,ρ,s) 6 0 for all sufficiently∈ ∞ large s R and so, since L(E,ρ,s) is continuous on [1, ), the intermediate≥ value theorem implies∈ that LT(E, ρ, 1) > 0. It thus follows that one has∞L (E,ρ)w R . ρ ∈ >0 L5. As in 2.5, let F/Q be a finite abelian extension over which ρ can be realised and let § τ be an F-linear representation of GQ such that C F τ ρ. By Corollary 23, the motive h1(E)(1) [τ] is critical and moreover, if L(E, ρ, 1)⊗ = 0≃ and Deligne’s period conjecture holds, then⊗ 6

(i) L (E,ρ) F×, (ii) L (E,ργ )=∈ L (E,ρ)γ for all γ Gal(F/Q). ∈ The result follows from this on noting that ργ ρ for all γ Gal(F/Q(ρ)). ≃ ∈ L6. (i) This follows directly from [16, Theorem 1]. (ii) In this case fρ is not divisible by any prime where E has bad reduction, so the claim follows from [16, Theorem 2a]. Corollary 25. Let E be an elliptic curve over Q. Suppose that L(E,ρ,s) has an analytic continuation to C for all Artin representations ρ over Q, and set BSD(E,ρ) = L (E,ρ). Then C1-C6 of Conjecture 4 hold subject to the following conditions: C1. The Birch–Swinnerton-Dyer conjecture holds for E over number fields. C2. Unconditional. C3. L(E,ρ,s) satisfies the functional equation and Conjecture 1. C4. L(E,ρ,s) satisfies the Riemann hypothesis. C5. L(E,ρ,s) satisfies Deligne’s period conjecture. C6. Stevens’s Manin constant conjecture holds for E/Q. In the following corollary we prove the remaining parts of Theorem 13. In particular, parts (1)-(4) record some of the formal consequences of the (conjectural) properties L1-L5 stated in Theorem 24 and parts (5)-(7) use the classical Birch Swinnerton-Dyer conjecture to make predictions about the norm of L (E,ρ). − Corollary 26. Let E/Q be an elliptic curve and let ρ be an Artin representation over Q. Suppose that L(E,ρ,s) admits an analytic continuation to all of C and that L(E, ρ, 1) =0. 6 14 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA

Suppose moreover that L(E,ρ,s) satisﬁes the functional equation, the Riemann Hypothesis and Deligne’s period conjecture. Then L (E,ρ ) (1) U(E,ρ) := ∗ is a root of unity in Q(ρ). L (E,ρ) (2) L (E,ρ) is invariant as a fractional ideal under complex conjugation. · OQ(ρ) (3) ζ L (E,ρ) Q(ρ, ζ)+ where ζ C is such that ζ2 = U(E,ρ). − · ∈2 1 ∈ d (ρ) dim ρ (4) U(E,ρ) = w w− and, if f is coprime to f , this equals ( 1) w det ρ(f ), ρ E,ρ E ρ − E E where det ρ is regarded as a primitive Dirichlet character.

Let G = Gal(Q(ρ)/Q) and choose number ﬁelds Fi, Fj′ and a positive integer m such that m g ⊕ ρ IndF /Q 1 IndF ′/Q 1 ⊕ i ≃ j g G i j M∈ M M (these exist by Remark 10), and write B for the unique positive real number such that

BSD(E/F ′) Bm = j j . Qi BSD(E/Fi) Suppose that the Birch Swinnerton-Dyer conjecture holds for E over number ﬁelds. Then − Q (5) N (L (E,ρ)) = B, with sign + if m is odd. Q(ρ)/Q ± (6) If ρ ρ∗ and ζ Q(ρ), then N + (ζ L (E,ρ)) = √B. 6≃ ∈ Q(ρ) /Q · ± (7) If ρ ρ∗ and ζ / Q(ρ), then N + (ζ L (E,ρ)) = B. 6≃ ∈ Q(ρ,ζ) /Q · ± Proof. 1) It follows from L5 that U(E,ρ) Q(ρ) and that ∈ U(E,ρ)g = L (E,ρg)/L (E,ρg) g G. ∀ ∈ In particular, one has U(E,ρ)g = 1 for all g G, and so U(E,ρ) is indeed a root of unity. | | ∈ 2) This follows directly from (1) on noting that, by L5, one has L (E,ρ∗)= L (E,ρ). 3) It follows from L2 and L5 that g g g g 2 2 g L (E,ρ (ρ )∗)= U(E,ρ )L (E,ρ ) = U(E,ρ)L (E,ρ) g G, ⊕ ∀ ∈ and so, recalling that wρg (ρg)∗ = 1 for any g G, it follows from L4 that ⊕ ∈ (U(E,ρ)L (E,ρ)2)g R g G. ∈ >0 ∀ ∈ Hence, taking ζ C such that ζ2 = U(E,ρ), we see that ζ L (E,ρ) Q(ρ, ζ)+. ∈ · ∈ 4) The ﬁrst statement follows immediately from L3 and the second from [5, Theorem 16]. g 5) Writing R(ρ)= g G ρ , it follows from L2 and L5 that ∈ L L L (E, R(ρ)) = NQ(ρ)/Q( (E,ρ)). m However, since R(ρ)⊕ IndF /Q 1 IndF ′/Q 1, another application of L2 gives ⊕ i i ≃ j j L ′ L L j (E, IndF /Q 1) L (E, R(ρ))m = j . L (E, Ind 1) Qi Fi/Q Hence, by L1 together with the Birch Swinnerton-Dyer conjecture, we get − Q L (E, R(ρ))m = Bm, and so the result follows on taking mth roots.

+ 6)&7) Let Γ = Gal(Q(ρ, ζ)/Q) and H = Gal(Q(ρ, ζ) /Q). If ρ ρ∗, then 6≃ γ η ρ (ρ ρ∗) , ≃ ⊕ γ Γ η H M∈ M∈ ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 15 and so it follows from L2, L3, L4 and L5 that

γ 2 L E, ρ = N + (L (E,ρ ρ∗)) = N + (ζ L (E,ρ)) . Q(ρ,ζ) /Q ⊕ Q(ρ,ζ) /Q · γ Γ M∈ On the other hand, L2 gives us L (E, R(ρ)) if ζ Q(ρ), L E, ργ = L (E, R(ρ))2 if ζ∈ / Q(ρ), γ Γ ∈ M∈ and so the results follow from L1 together with the Birch Swinnerton-Dyer conjecture. − 3. Arithmetic applications In order to obtain arithmetic applications of Conjecture 4 we shall make use of (C5), which is the analogue of the Galois equivariance property of L-values. As we do not have an exact expression for BSD(E,ρ) in general, we shall take the following approach. The g representation τ = g Gal(Q(ρ)/Q) ρ has rational trace and hence BSD(E, τ) can, on the one hand, be expressed∈ in terms of BSD-quotients BSD(E/K ) for suitable fields K (see L i i Remark 10), and, on the other hand, is the norm of BSD(E,ρ) from Q(ρ) to Q by (C5). As we shall illustrate, this places non-trivial constraints on the BSD(E/Ki), and hence on ranks and the Tate–Shafarevich groups. We stress that the expression is the norm of an element of Q(ρ), rather than just of a fractional ideal. Theorem 27. Let G be a finite group and ρ an irreducible representation. Write m g ⊕ G G ρ = IndH 1 IndH′ 1 , i ⊖ j g Gal(Q(ρ)/Q) i j ∈ M M M for some m Z and some subgroups H ,H′ < G. If Conjecture 4 (C1, C2, C5) hold, then ∈ i j for every elliptic curve E/Q and Galois extension F/Q with Galois group G, either ρ, E(F ) > 0 h Ci or Hi i BSD(E/F ) m ′ = NQ(ρ)/Q(x) Hj Qj BSD(E/F ) for some x Q(ρ). Moreover, if ρ is non-trivial, G is abelian of exponent d and (C6) of ∈ Q Conjecture 4 holds, then one can take x Z[ζd] provided that either fE and fρ are coprime, or E is semistable and has no non-trivial∈ isogenies. Proof. If ρ, E(F ) = 0, then the formula is satisfied by x = BSD(E,ρ). h Ci In order to make use of the above theorem in specific settings, we will need to control the various terms in the BSD factors of the formula. For convenience of the reader we have recorded in 3.4 some standard facts about Selmer groups, Tamagawa numbers and the term ω/ωmin that§ will be used in our computations. | | 3.1. Interplay between p-primary parts of the Tate-Shafarevich group. For our first application we will take the simplest setting, when the Galois group is cyclic of prime order, and make use of the fact that the ratio of BSD-terms is the norm of a principal ideal. The basic idea is that if the p-part of this number cannot be expressed as the norm of a principal ideal, then, necessarily, the q-primary part must be non-trivial for some other prime q. 16 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA

Theorem 28. Let ℓ and p be primes such that the primes above p in Q(ζℓ) are non-principal and have residue degree 2. If Conjecture 4 holds then for every semistable elliptic curve E/Q X with no non-trivial isogenies, with E/Q[p] =1 and cv =1 for all rational primes v, and for every cyclic extension F/Q of degree| ℓ with| E(F )= E(Q), 2 if X [p∞] = p then X [q∞] =1 for some q = p, ℓ. | E/F | | E/F | 6 6 X Proof. We will in fact prove the stronger statement that E/Q[q∞] is strictly smaller than X E/F [q∞] for some q = p,ℓ. The fact that it is a subgroup for all q = ℓ is standard: it is k 6 6 true for q -Selmer groups by Lemma 36, and as E(Q)= E(F ) it is also true for X[q∞]. A 1-dimensional faithful representation χ of Gal(F/Q) C has Q(χ)= Q(ζ ). Now ≃ ℓ ℓ χg = C[G] 1, ⊖ g Gal(Q(χ)/Q) ∈ X so by Theorem 27 BSD(E/F ) = N (x) BSD(E/Q) Q(ζℓ)/Q for some x Z[ζℓ]. As E(F ) = E(Q), it follows that E(F )tors = E(Q)tors and RegE/F = rk E/Q ∈ ℓ RegE/Q (Lemma 36(2)). As E/Q is semistable, the contributions to CE/F and CE/Q only come from Tamagawa numbers (Lemma 36(5)). These are trivial over Q by hypothesis, so are also trivial at all primes that split in F/Q; if v is a prime of multiplicative reduction that does not split then the corresponding Tamagawa number over F is 1 unless the reduction is split multiplicative and the prime ramifies, in which case it is ℓ (Lemma 36(4)). Putting this together we deduce that X | E/F | ℓn = N (x) forsome n Z. X · Q(ζℓ)/Q ∈ | E/Q| Note that ℓ is totally ramified in Q(ζℓ) and the ideal above it (ζℓ 1) is principal. Thus if X X 2 − E/F [q∞] = E/Q[q∞] for all q = p,ℓ, it would follow that p is the norm of a principal 6 2 ideal of Z[ζℓ]. By assumption, the primes above p have norm p and are non-principal, so this is not the case. Example 29. Let E/Q be a semistable elliptic curve with no non-trivial isogenies, with X v cv = 1 and E/Q =1, and let F/Q be the degree 229 subfield of Q(ζ2749). In this setting, 2 Q if E(Q)= E(F ) then X(E/F ) = p , | | 6 for the prime p = 1148663. Indeed, p is a prime which has residue degree 2 in Q(ζ229), and the prime above it is non-principal (this is hard to achieve, which is why ℓ = 229 is so large), so this is a consequence of the above theorem. However, it is perfectly possible for such a curve to have E(Q)= E(F ) and X(E/F ) = p2 (integer coprime to p), | | × as, for instance, is the case for the elliptic curve 2749a1. (This is based on a Magma compu- tation of the analytic order of X and the analytic rank, and assumes the BSD conjecture.)

3.2. Forcing non-trivial Selmer groups. The BSD-terms BSD(E/F Hi ) in Theorem 27 are composed of “hard” global invariants (Tate–Shafarevich group and points of inﬁnite order) and “easy” local invariants (Tamagawa numbers and diﬀerentals). We will now illustrate how the result can be used to make the easy local data force non-trivial behaviour of global invariants. Once again, we will exploit the fact that the ratio of BSD-terms is the norm of a principal ideal. We focus on non-abelian groups of the form G = Cq1q2 ⋊Cr, and begin by simplifying the norm condition. ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 17

Theorem 30. Let q1, q2, r be distinct odd primes with q1, q2 1 mod r, but q1, q2 1 2 th ≡ 6≡ mod r , and with q1 an r power in Z/q2Z and vice versa. Let G = Cq1q2 ⋊Cr with Cr acting non-trivially on both the Cq1 and Cq2 subgroups. If Conjecture 4 holds, then for every elliptic curve E/Q and every Galois extension F/Q with Galois group G, either rk E/F > 0 or

Cr BSD(E/F ) BSD(E/Q) a b C ⋊C C ⋊C = q1 q2m BSD(E/F q1 r ) BSD(E/F q2 r ) for some a,b Z and m Q with m 1 mod q q . ∈ ∈ ≡ 1 2 G Proof. Let ψ be a faithful 1-dimensional representation of Cq1q2 and ρ = Ind ψ. This Cq1q2 Cr is a faithful r-dimensional irreducible representation of G, and Q(ρ) = Q(ζq1q2 ) with the Cr action coming from Cr Aut(Cq1q2 ) = (Z/q1q2Z)× = Gal(Q(ζq1 q2 )/Q). Applying Theorem 27 to the identity ⊂ g G G G ρ = IndC 1 1 IndC ⋊C 1 IndC ⋊C 1 r ⊕ ⊖ q1 r ⊖ q2 r g Gal(Q(ρ)/Q) ∈ M shows that either rk E/F > 0 or the ratio of the BSD terms in the statement is a norm of an element x Q(ρ). It thus suffices∈ to show that the norm of every non-zero principal ideal (x) in Q(ρ) is of the form qaqbm, for some a,b Z and m Q with m 1 mod q q . Observe that 1 2 ∈ ∈ ≡ 1 2 Q(ζq1q2 )= Q(ρ)(ζq1 )= Q(ρ)(ζq2 ). In particular, Q(ζq1q2 )/Q(ρ) is unramified at all primes, so by class field theory

ordp x Frobp = id Gal(Q(ζq1q2 )/Q(ρ)), p ∈ Y the product taken over the primes of Q(ρ) (all the inﬁnite places being complex as r is odd). th 2 By hypothesis q1 is an r power in Fq and r ∤ F× , so Fq (ζq )/Fq has degree coprime to 2 | q2 | 2 1 2 r; hence every prime above q must split in Q(ρ)(ζ )/Q(ρ), that is Frobp =id for every p q . 2 q1 | 2 Similarly Frob =id for every p q , and hence Frobordp x = id. For a prime p ∤ q , q p 1 p q1,q2 p 1 2 | N(p) | the Frobenius element is simply ζ ζq q , which shows that q1 q2 7→ 1 2 Q ordp x N(p) =1 (Z/q q Z)×, ∈ 1 2 p∤Yq1,q2 and hence the norm of x is of the required form.

Corollary 31. Let F/Q be a Galois extension of degree 3q1q2 that contains a Galois cubic ﬁeld K/Q and ℓ a prime that satisfy

q1, q2 4 or 7 mod 9, and q1 is a cube modulo q2 and vice versa, • q1q2 ≡ q1 q2 4 4 4 =0mod3, • ⌊3 ⌋−⌊ ⌋−⌊ ⌋ 6 ℓ 1 mod q1, but ℓ = 1 mod q1 and ℓ = 1 mod q2, • ℓ has≡ residue degree 36 and ramiﬁcation6 degree q q in F/Q. • 1 2 If Conjecture 4 holds, then every elliptic curve E/Q that has additive reduction at ℓ of Kodaira type III and good reduction at other primes that ramify in F/Q, must have a non- trivial p-Selmer group for some prime p ∤ [F :Q].

Proof. First note that Gal(F/Q) Cq1q2 ⋊C3 (the inertia group at ℓ is tame, so Cq1q2 is a subgroup, and the extension is split≃ by the Schur-Zassenhaus theorem). As ℓ = 1 mod q , 6 1 there is no Galois extension of Q of degree q1 that is ramiﬁed at ℓ, and similarly for q2. It follows that C3 must act non-trivially both on Cq1 and Cq2 . Moreover, q1 and q2 are both cubes modulo each other and are 4 or 7 mod 9, so Theorem 30 applies with r = 3. ≡ 18 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA

If the Selmer group Selp(E/F ) is trivial for a prime p ∤ [F :Q], then it is also trivial over any subﬁeld of F (see Lemma 36(1)), and hence neither the torsion nor X contribute to the p-part of the BSD quotient in the theorem. The rank over F is then also 0, so all the regulators are trivial. Thus by Theorem 30, supposing that Selp(E/F ) = 0 for all p ∤ [F :Q],

CE/F Cr CE/Q a b · = q1 q2m C Cq ⋊Cr C Cq ⋊Cr E/F 1 · E/F 2 for some a,b Z and m 1 mod q1q2. However, in our setup this is not the case, as we now explain. ∈ ≡ First observe that all primes of good reduction, which include all ramified primes in F/Q apart from ℓ, have trivial Tamagawa numbers and ω/ωmin terms in all extensions, and hence do not contribute to the ratio of the C-terms| above.| If q = ℓ is a prime of bad reduction for E/Q, then, by assumption, it is unramified in F/Q. The6 minimal differential at q then remains minimal in all extensions, so that the ω/ωmin terms for these primes are | | all 1. Moreover, the decomposition group at q is either trivial or cyclic of order 3, q1, q2 or q1q2, and a straightforward case-by-case check shows that the Tamagawa numbers from the primes above q contribute a perfect cube to the above ratio of C terms (in fact each extension of Qq always appears in the expression a multiple of 3 times). Finally consider the primes above ℓ. As ℓ has ramification degree q1q2 in F/Q, it is totally C C ⋊C C ⋊C ramified in F r , F q1 r and F q2 r . Thus E has reduction type III or III∗ at the prime above ℓ in these fields, and the corresponding Tamagawa number is always 2 (see [11] IV.9). § The minimal model over Qℓ does not remain minimal (valuation of the discriminant goes · · · min 3 q1q2 3 q1 3 q2 n above 12, Lemma 36(6)) and the ω/ω terms contribute ℓ⌊ 12 ⌋/ℓ⌊ 12 ⌋ℓ⌊ 12 ⌋ = ℓ , where n = 0 mod 3 by assumption. | | Putting6 these computations together shows that

C Cr CE/Q E/F · = x3ℓn C Cq ⋊Cr C Cq ⋊Cr E/F 1 · E/F 2 for some x Q and integer n = 0 mod 3. However, ℓ has order 3 in Fq×1 and so, as q = 1 mod 9,∈ it is not a cube in F6 . As q is a cube mod q , this expression cannot be of 1 6 q1 2 1 the form qaqbm for any m 1 mod q q , which gives the desired contradiction. 1 2 ≡ 1 2 Remark 32. Number fields satisfying the hypotheses of Corollary 31 do exist. For example, + we can take q1 = 643 and q2 = 43. For simplicity, let us take K = Q(ζ9) (class number 1) and then choose ℓ and F using class field theory as follows. First pick five candidate primes ℓ1,...ℓ5 that satisfy the 3rd and 4th bullet points of the corollary — these are congruence 3 conditions, so such primes exist. As ℓi has residue degree 3 in K and ℓi 1 mod q1, the 5 ≡ group ( K / ℓi)× has a Cq×1 quotient. The unit group of K has rank 2, so the ray class O n group (( K / ℓi)×/image of × ) has a Gal(K/Q)-stable C× quotient for some n 3. In O Q OK q1 ≥ particular, as q1 1 mod 3 and Fq contains the 3rd roots of unity, there are at least three Q ≡ 1 Gal(K/Q)-stable Cq1 -quotients whose corresponding fields F1, F2, F3 under global class field theory are linearly disjoint. By construction, the Fi are Galois over Q, have degree q1 over K and only the ℓi can ramify in Fi/K. As K has class number 1 and the Fi are linearly disjoint, at least three of the ℓi must ramify in some of the fields. Now repeating the same construction with the same ℓi for q2 similarly yields three fields F1′, F2′, F3′ of degree q2 over K. One of the ℓi must ramify both in one of the Fi and in one of the Fi′, say ℓ1 ramifies in F1 and in F2. We can the take F = F1F2 and ℓ = ℓ1. 3.3. Forcing points of infinite order. For our final type of application of Theorem 27, we will make the local data force the existence of points of infinite order on elliptic curves. ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 19

This time, the idea is to make sure that the ratio of BSD-terms in the theorem cannot be the norm of an element at all, and hence E/F must have positive rank. In order to do this, we need a way of controlling the Tate–Shafarevich group. In general, this is very difficult, so we will simply make use of the fact that it has square order and that all squares are norms from quadratic fields. Theorem 33. Suppose Conjecture 4 holds. Let E/Q be an elliptic curve, F/Q a Galois extension with Galois group G, ρ an irreducible representation of G and m g ⊕ ρ = IndF /Q 1 IndF ′/Q 1 i ⊖ j g Gal(Q(ρ)/Q) i j ∈ M M M Qi CE/Fi for some m Z and subfields Fi, Fj′ F . If either ′ is not a norm from some Q CE/F ∈ ⊆ j j quadratic subfield Q(√D) Q(ρ), or if it is not a rational square when m is even, then E has a point of infinite order⊂ over F .

Q BSD(E/Fi) Proof. Suppose rk E/F = 0. By Theorem 27, ′ is the m-th power of the norm of Q BSD(E/Fj ) an element of Q(ρ). In particular it is a norm from Q(√D), and if m is even it is a rational square. As the rank is zero over F , the regulators that enter the BSD-terms are all 1. The contributions from X and torsion are all squares, and hence automatically norms from Q(√D). Q C i E/Fi √ It follows that the remaining expression ′ must be a norm from Q( D) as well, and Q CE/F j j a rational square in case m is even. The criterion of Theorem 33 can be applied in many Galois groups to find local conditions on elliptic curves that guarantee the existence of points of infinite order. We illustrate it on the group of quaternions, Q8: Corollary 34. Suppose Conjecture 4 holds. Let F/Q be a Galois extension with Galois group Q8. Then every elliptic curve E/Q with good reduction at 2 and 3 and with an odd number of potentially multiplicative primes that do not split in F/Q must have a point of infinite order over F .

Proof. Let ρ be the 2-dimensional irreducible representation of Q8, so that

2 Q8 Q8 ρ⊕ = Ind 1 Ind 1. 1 ⊖ C2 C We will show that E/F has odd 2-adic valuation, where L = F C2 . The result then follows CE/L from the theorem. Observe that if a prime p splits in F/Q, then it necessarily already splits in L/Q. Indeed, if there is only one prime above p in L, then the decomposition group at p surjects onto Q8/C2. The only subgroup with this property is the whole of Q8, so there is only one prime above p in F . It follows that split primes contribute square contributions to CE/F . CE/L As E has good reduction at 2, these primes do not contribute to the ratio: ω remains minimal in all ﬁeld extensions of Q2 and the local Tamagawa number is always 1 (Lemma 36). At primes v ∤ 2, the contribution from ord ω/ωmin will clearly be zero. Thus 2 | |v C ord E/F = ord (c /c ), 2 C 2 w v E/L p B X∈ where B is the set of primes of bad reduction of E that do not split in F/Q, and where v and w are the primes above p in L and F , respectively. 20 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA

If p B then p necessarily has residue degree 2 and ramification degree 4 in F/Q and the prime∈ above it ramifies in F/L, as the only possible choice for the (tame!) inertia subgroup and its cyclic quotient is Q8/C4 = C2. In particular, if p is a prime of potentially multiplicative reduction then E has split multiplicative reduction at v in L and cw = 2cv (Lemma 36). If p is a prime of potentially good reduction then the p-adic valuation δ of the minimal discriminant of E/Qp determines the Kodaira type of E at v and at w. Recall that the Tamagawa number of E over a local field M is the number of Frobenius invariant nr nr points of E(M )/E0(M ), so we read off from [11] Chapter 9 Table 4.1 that the pair of Tamagawa numbers cv,cw is either 3, 3 or 3, 1 (δ =2, 10), 1, 1 or 4, 1 (δ =3, 9, noting that Lv is a quadratic unramified extension of a field, so Frobenius has odd order on E/E0 over Lv), 2,2 (δ =4, 8), or 1,1 (δ = 6). Thus in all cases of potentially good reduction ord2 cw/cv is even. The result follows. As a final application, we will prove a result on the Birch–Swinnerton-Dyer conjecture in dihedral extensions. This time we will classify the cases when our local BSD-term data predicts that ρ appears in E(F )C and compare it to the corresponding root number predictions. Theorem 35. Suppose Conjecture 4 holds. Let F/Q be a Galois extension with Galois group D2pq, with p, q 3 mod 4 primes, and let ρ be a faithful irreducible Artin representation that factors through≡ F/Q. Then for every semistable elliptic curve E/Q, if ord L(E,ρ,s) is odd, then ρ, E(F ) > 0. s=1 h Ci Proof. We first remark that this L-function does have an analytic continuation to C and satisfies the standard functional equation. (It can be expressed as a classical Rankin–Selberg product. Alternatively, ρ is induced from a 1-dimensional representation ψ of Gal(F/K), where K is the quadratic subfield of F , and so L(E,ρ,s)= L(E/K,ψ,s)= L(π ψ,s), E/K ⊗ where πE/K is the automorphic form obtained by cyclic base change from the modular form attached to E/Q by modularity.) We apply Theorem 27 to the identity ρg = IndG 1 IndG 1 IndG 1 1, C2 ⊖ D2p ⊖ D2q ⊕ g Gal(Q(ρ)/Q) ∈ M + where G = D2pq. Here Q(ρ)= Q(ζpq) contains the quadratic field Q(√pq). Since squares are always norms from quadratic fields we deduce that either ρ, E(F ) > 0 or h Ci CE/Lpq CE/Q RegE/Lpq RegE/Q = NQ(√pq)/Q(x) for some x Q(√pq), CE/Lp CE/Lq · RegE/Lp RegE/Lq ∈

C2 D2q D2p where Lpq = F , Lp = F and Lq = F are the intermediate ﬁelds of degree pq,p and q over Q, respectively. Write M for the set of primes of multiplicative reduction of E/Q, sr = 1 if the reduction at a prime r M is split and sr = 1 if it is non-split, and write er and fr for the ramiﬁcation and∈ residue degree of a prime− r in F/Q, respectively. Set X = v in K r M,e =1,f =2 r M,e =2,s = 1 . { |∞ }∪{ ∈ r r }∪{ ∈ r r − } C C Reg Reg E/Lpq E/Q E/Lpq E/Q #X Claim 1: If ρ, E(F )C = 0 then C C Reg Reg = (pq) , where “ ” is h i E/Lp E/Lq · E/Lp E/Lq · shorthand for a rational square. Claim 2: ord L(E,ρ,s) #X mod 2. s=1 ≡ Observe that pq is not the norm of any element of Q(√pq). Indeed, as p 3 mod 4, the 2 2 ≡ norm equation pq = a pqb is not even soluble in Qp. It thus follows from the two claims and the formula above− that ρ, E(F ) > 0, which proves the theorem. h Ci ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 21

Proof of Claim 2: The parity of the order of vanishing of the L-function is given by the root number wE,ρ. As E/Q is semistable and dim ρ = 2, by [5] Thm. 1,

− Ir w = ( 1)dim ρ sdim ρ det(Frob ρIr ), E,ρ − r r | r M Y∈ where dim ρ− is 1 or 2 according to whether K is complex or real, Frobr is any choice of Frobenius element at r in Gal(F/Q), and ρIr is the subspace of ρ that is pointwise fixed by the inertia subgroup Ir. If a prime r M is unramified in F/Q then its contribution to the product is 1 if and only if Frob has∈ order 2. If it ramifies, then the contribution is 1 − r − if and only if Ir has order 2 (in D2pq, Ir = 2 forces Frobr to be trivial) and sr = 1. In #X | | − other words wE,ρ = ( 1) . Proof of Claim 1: Since− E/Q is semistable, its global minimal differential remains minimal in all field extensions, so we can write CE/L = r M Cv r(L) with Cv r(L)= v r cv(E/L). | | The group D has five rational irreducible representations:∈ trivial 1, sign| ǫ, τ that 2pq Q Q p factors through the D2p-quotient and similarly τq and τpq. Now pick points P1,...,Pa E(Q) that form a basis for E(Q) Q, the 1-isotypical component of E(F ) Q. Complete∈ ⊗Z ⊗ it to a basis P1,...,Pa,Q1,...Qb E(Lp) for E(Lp) Q, with the Qi belonging to the τ -isotypical component of E(F ) ∈Q; and similarly to P⊗ ,...,P , R ,...R for E(L ) Q p ⊗ 1 a 1 b q ⊗ with the Ri belonging to the τq-isotypical component. By assumption, τpq does not appear in E(F ) Q, so that the P ,Q and R together form a basis for E(L ) Q. Moreover, as ⊗ i i i pq ⊗ the height pairing on E(F ) is Galois invariant, the spaces spanned by the Pi, the Qi and the Ri are orthogonal to each other. Finally, recall that the height pairing scales under field extensions by the degree, so that the ratio of the regulators is Reg Reg E/Lpq E/Q rk E/L rk rk E/L rk = q p− E/Q p q− E/Q , RegE/Lp RegE/Lq · the square error coming from the fact that our bases span finite index sublattices of E(Q), E(Lp), E(Lq) and E(Lpq) (see Lemma 36). X As the p- and q-primary parts of E/F are finite (which is therefore also true over all the subfields), the known cases of the parity conjecture for E/Q, E/Lp and E/Lq ([6] Thm. 1.3), tell us that the parity of each exponent in the above formula is determined by the corresponding root number. Thus, like for the C terms, we can express this as a product

rk E/L rk E/Q θ(q) q p− = q r , · r M ∈ Y∪{∞} (q) where θr is 0 or 1 depending on whether w(E/Q ) w(E/(L ) ) is 1 or 1; and similarly r v r p v − for the exponent of p. Hence | Q

CE/Lpq CE/Q RegE/Lpq RegE/Q = Zr, CE/L CE/L · Reg Reg · p q E/Lp E/Lq r M ∈ Y∪{∞} θq θp where Zr = Cv r(Q)Cv r(Lp)Cv r(Lq)Cv r(Lpq)q r p r . Thus it now suﬃces to check that Z = pq for r | X and| Z = | for r / |X. r ∈ r ∈ To explicitly determine Zr, we systematically work through all possibilities. Recall that the local root number w(E/Lv) is +1 for good and non-split multiplicative reduction and 1 for split multiplicative reduction and for archimedean places. Recall also that if the Kodaira− type of E/Lv is In then the Tamagawa number is n if the reduction is split, and 1 or 2 if it is non-split, depending on whether n is odd or even, which we will denote byn ˜. Finally, multiplicative reduction of type In becomes of type Ien after a ramiﬁed extension of degree e, split reduction remains split, and non-split reduction becomes split if the extension has 22 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA

even residue degree. We now tabulate the contribution to the above product from a prime r depending on its ramiﬁcation degree er and residue degree fr in F/Q; in D2pq these uniquely determine the inertia and decomposition subgroups, and hence the splitting behaviour of r in all intermediate extension. The values are constrained by the fact that both the (tame!) inertia group is cyclic of order er and normal in the decomposition group with a cyclic quotient of order fr. The entries for split and non-split multiplicative reduction of type In are separated by a “;”.

θ(q) θ(p) er, fr Cv r(Q) Cv r(Lp) Cv r(Lq) Cv r(Lpq) q r p r Zr 1, 1 n| ;ñ np;| n ˜p nq;| n ˜q npq|;n ˜pq 1;1 1;1 p+1 p−1 q+1 q−1 pq+1 pq−1 1, 2 n;ñ n 2 ;nn ˜ 2 n 2 ;nn ˜ 2 n 2 ;nn ˜ 2 q; q p; p pq 1, p n;ñ n;n ˜ nq;n ˜q nq;n ˜q 1;1 1;1 1, q n;ñ n;n ˜ np;n ˜p np;n ˜p 1;1 1;1 1, pq n;ñ n;n ˜ n;n ˜ n;n ˜ 1;1 1;1 p−1 p−1 q−1 q−1 pq−1 pq−1 2, 1 n;ñ (2n) 2 n; 2 2 n˜ (2n) 2 n; 2 2 n˜ (2n) 2 n; 2 2 n˜ q;1 p;1 pq; p, 1 n;ñ np;n ˜ nq;n ˜q (np)q;n ˜q 1;1 1;1 q+1 q−1 q+1 q−1 p, 2 n;ñ np;n ˜ n 2 ;nn ˜ 2 (np) 2 ;n ˜(np) 2 1;1 p; p p, q n;ñ np;n ˜ n;n ˜ np;n ˜ 1;1 1;1 q, 1 n;ñ nq;n ˜ np;n ˜p (nq)p;n ˜p 1;1 1;1 p+1 p−1 p+1 p−1 q, 2 n;ñ nq;n ˜ n 2 ;nn ˜ 2 (nq) 2 ;n ˜(nq) 2 1;1 q; q q, p n;ñ nq;n ˜ n;n ˜ nq;n ˜ 1;1 1;1 q−1 q−1 q−1 q−1 2p, 1 n;ñ np;n ˜ (2n) 2 n; 2 2 n˜ (2np) 2 np; 2 2 n˜ 1;1 p;1 p−1 p−1 p−1 p−1 2q, 1 n;ñ np;n ˜ (2n) 2 n; 2 2 n˜ (2nq) 2 nq; 2 2 n˜ 1;1 q;1 pq, 1 n;ñ np;n ˜ nq;n ˜ npq;n ˜ 1;1 1;1 pq, 2 n;ñ np;n ˜ nq;n ˜ npq;n ˜ 1;1 1;1

Finally, note that if the quadratic field K is real, then F/Q is totally real, so Lp has p q θ∞ infinite places and q = 1, and similarly for Lq; hence Z = 1. If K is imaginary, then q ∞ p 1 θ∞ Lp has one real and −2 (=odd) complex places and q = q, and similarly for Lq; hence Z = pq. Thus, indeed, Zr = pq for r X and Zr = for r / X, as required. ∞ ∈ ∈ 3.4. Summary of some basic properties. We list some standard results regarding elliptic curves over local and global fields. We give brief proofs as, while these results are well-known, they may not always be easy to find in the literature. Lemma 36. Let E/K be an elliptic curve over a number field, F/K a field extension of finite degree d. Let v be a finite place of K with w v a place above it in F , and ω and ω | v w minimal differentials for E/Kv and E/Fw, respectively.

(1) If F/K is Galois, then Seln(E/K) is a subgroup of Seln(E/F ) for all n coprime to d. (2) For P,Q E(K), their N´eron–Tate height pairings over K and F are related by ∈ P,Q F = d P,Q K . h i h i drk E/K (3) If rk E/F = rk E/K, then RegE/F = n2 RegE/K , where n is the index of E(K) in E(F ). (4) If E/Kv has good reduction then cv = 1. If E/Kv has multiplicative reduction of min Kodaira type In then n = ordv ∆E,v and cv = n if the reduction is split, and cv =1 (respectively, 2) if the reduction is non-split and n is odd (respectively, even). (5) If E/K has good or multiplicative reduction then ω /ω =1. v | v w|w (6) If E/Kv has potentially good reduction and the residue characteristic is not 2 or 3, min eF/K ordv ∆E,v then ω /ω = q⌊ 12 ⌋, where q is the size of the residue ﬁeld at w. | v w|w ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 23

(7) If v has odd residue characteristic, E/Kv has potentially multiplicative reduction and Fw/Kv has even ramification degree, then E/Fw has multiplicative reduction. (8) Multiplicative reduction becomes split after a quadratic unramified extension. Proof. (1) In the inflation-restiction sequence H1(Gal(F/K), E(F )[n]) H1(K, E[n]) H1(F, E[n]), the first term is killed both by Gal(F/K) and by n, and→ is therefore trivial.→ Thus the second map and its restriction to n-Selmer| groups| are injective. (2) This follows from the definition of the height pairing, see [13] (1.6). (Note that it is not normalised as for the absolute height.) (3) Follows from (2) and the fact that the height pairing is bilinear and non-degenerate. (4) [11] IV.9. § (5) As E/Kv has good or multiplicative reduction, its minimal Weierstrass model over Kv remains minimal over Fw, so ωv is also a minimal differential over F . min (6) In this setting ordw ∆E,w < 12, so the result follows from the formula in Notation 17. (7) This follows from the theory of the Tate curve, see e.g. [11] Exc. 5.11. (8) Clear from the definition of non-split multiplicative reduction.

4. Arithmetically similar twists with different L-values In this section we discuss the problem of formulating a precise Birch–Swinnerton-Dyer type formula for twists of elliptic curves by Dirichlet characters χ. We will make the information that we know about L (E,χ) explicit and discuss the difficulties illustrated in Example 3. We will also give many numerical examples, for the benefit of those readers who may wish to analyse these L-values in more detail. The numerical examples throughout this section were worked out using Magma [2]. The orders of X given are strictly speaking “analytic orders of X”, that is the orders that are predicted by the Birch–Swinnerton-Dyer conjecture. Notation 37. Recall from Notation 15 that we identify Dirichlet characters χ with their corresponding 1-dimensional Galois representations. We write Kχ for the abelian number field cut out by the kernel of χ, that is for the smallest extension Kχ/Q such that χ factors through Gal(Kχ/Q). In the context of Dirichlet characters, we already know from Theorem 13 a substantial amount about L(E, χ, 1) in terms of arithmetic data: Theorem 38. Suppose Stevens’s Manin constant conjecture holds for E/Q. Let χ be a non-trivial primitive Dirichlet character of order d and conductor coprime to fE. Then L (E,χ) Z[ζ ] and, if L(E, χ, 1) =0, then furthermore ∈ d 6 d+1 ζ L (E,χ) R, for ζ = χ(f ) 2 χ( 1)w . · ∈ E − E If rk E/Q =0 and the Birch–Swinnerton-Dyer conjecture holdsp for E over Q and Kχ, then X χ E(Q)tors E/Kχ v cv(E/K ) N + (ζ L (E,χ)) = | | | | . Q(ζd) /Q · ± E(Kχ) X c (E/Q) | tors|s | E/Q| Qp p χ 1 If moreover d is odd and BSD(E/K ) = BSD(E/Q) then L (E,χQ ) = ζ− u for some unit u × + . ∈OQ(ζd) Proof. The first claim follows from Theorem 13 (8,9,10) with ρ = χ. g Applying Theorem 13(12) with the identity g Gal(Q(ζ )/Q) χ 1 = C[G] shows that ∈ d ⊕ L BSD(E/Kχ) NQ(ζ )+/Q(ζ (E,χ)) = . SinceL the conductor of E is coprime to that of χ, d · ± BSD(E/Q) the primes of bad reductionq of E are unramified in Kχ/Q, so a global minimal differential 24 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA for E/Q remains minimal over Kχ and hence all the contributions of the form ω/ωmin to the BSD-terms are trivial. This proves the desired second formula. | | For the final claim, note that as rk E/Q=0 and d is odd, we must have w =χ( 1)=1, E − so that ζ Q(ζd). The result now follows from the previous parts. ∈

Under the above assumptions, we can predict the value L(E, χ, 1) from Birch–Swinnerton- + Dyer type information up to an element of norm 1 in Q(ζd) . In fact, since L (E,χ) is integral, the prediction is stronger than that. For instance,± if χ has order 3 and BSD(E/Kχ)= BSD(E/Q) then L(E, χ, 1) is fully determined up to a sign. However, this final ambiguity appears to be severe: Theorem 39. For elliptic curves E/Q and Dirichlet characters χ as in Theorem 38, L X (1) (E,χ) cannot be expressed purely as a function of χ, of E(Q), E/Q, p E(Qp)/E0(Qp) χ χ χ χ as abelian groups and of E(K ), X χ , E(K )/E (K ) as Gal(K /Q)-modules. E/K v v 0 v Q (2) The fractional ideal (L (E,χ)) cannot be expressed purely as a function of χ, and of E(Q), X χ X Q χ χ E/Q, p E(Qp)/E0(Qp), E(K ), E/Kχ and v E(Kv )/E0(Kv ) as abelian groups. χ Here the productsQ are taken over all primes of Q andQ of K , and E0 denotes the usual subgroup of points of non-singular reduction. This theorem follows from the fact that one can find curves with identical arithmetic invariants listed in (1) and (2), but with different modified L-values L (E,χ). This is shown by the next two examples, where most of the objects listed in (1) and (2) are trivial.

Example 40. Let E1/Q be the elliptic curve given by y2 + y = x3 8x 9, − − and E2/Q be another elliptic curve given by y2 + y = x3 + x 1, − which have Cremona labels 307a1 and 307c1, respectively. Let χ be the primitive Dirichlet character of order 5 and conductor 11 deﬁned by χ(2) = ζ5. Both curves have

χ χ E (Q) = E (K ) = X = X χ = c (E /Q)= c (E /K )=1. | i | | i | | Ei/Q| | Ei/K | p i v i p v Y Y In particular, all the groups listed in Theorem 39 are trivial. In fact, the curves also have the same conductor fEi = 307 and the same discriminant ∆Ei = 307. However, their modiﬁed L-values diﬀer: − L L 4 2 (E1,χ)=1, (E2,χ)= ζ5 (1 + ζ5) . Remark 41. As the discriminants for the two curves in the above example are the same and thus in particular have the same sign, both curves have the same number of connected components over R. In other words, one can add the group of real connected components E(R)/E0(R) to the list of groups in Theorem 39(1), as well as the conductor and the discriminant of E.

Example 42. Let E1/Q be the elliptic curve given by y2 + y = x3 x2 1, − − and E2/Q be another elliptic curve given by y2 + xy = x3 + x2 3x 4, − − ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 25 which have Cremona labels 291d1 and 139a1, respectively. Let χ be the primitive character 3 of order ﬁve and conductor 31 deﬁned by χ(3) = ζ5 . Both curves have

χ χ 2 E (Q) = E (K ) = X = c (E /Q)= c (E /K ) = 1 and X χ = 11 . | i | | i | | Ei/Q| p i v i | Ei/K | p v Y Y

The discriminants ∆E1 = 291 and ∆E2 = 139 again have the same sign. For these curves, − − L (E ,χ)=2ζ3 ζ2 ζ + 2 and L (E ,χ)=5ζ3 + ζ2 + ζ +5. 1 5 − 5 − 5 2 5 5 5 These factorise as

(L (E1,χ)) = p1p2 and (L (E2,χ)) = p3p4, where p1 = (11, 7+ ζ5), p2 = (11, 8+ ζ5), p3 = (11, 6+ ζ5), p4 = (11, 2+ ζ5) are the primes of Q(ζ5) above 11. We note that it is plausible that the exact factorisation can be recovered from the Galois module structure of X. Unfortunately, it appears to be beyond our computational reach to check this at present. (See, however, the recent work of Burns and Castillo [3] Rmk. 7.4.) Remark 43. In the above example, our results on L-values are strong enough to predict that the ideal L (Ei,χ) must be either p1p2 or p3p4, though, as the example illustrates, they do not allow us decide which of the two occurs. To see why the factorisation must be one of these two, consider any Dirichlet character χ of order 5 and any elliptic curve E/Q satisfying χ BSD(E/K ) 2 the conditions of Theorem 38 and additionally BSD(E/Q) = 11 . Then by Theorem 13 (10), 2 (11) and (6), L (E,χ) is an element of Z[ζ5] of norm 11 and generates an ideal that is fixed by complex conjugation. Hence (L (E,χ)) must be either p1p2 or p3p4. For those who may be interested in investigating these L-values further, we end by giving a range of further examples. All elliptic curves below are given by their Cremona labels. Example 44. There are plenty of curves that have trivial Mordell–Weil groups, X and Tamagawa numbers both over Q and over Kχ for the same Dirichlet character χ of order 5 as in Example 40. Here we have chosen some groups of such curves that also have the same conductors, but, as in the example, have different modified L-values (here u =1+ ζ5 is a fundamental unit in Q(ζ5)).

E L (E,χ) E L (E,χ) E L (E,χ) E L (E,χ) E L (E,χ) 2 3 −1 2 2 307a1 1 432g1 u 714b1 1 1187a1 ζ5 u 1216g1 −ζ5 u 4 2 4 −1 3 4 −3 −1 307c1 ζ5 u 432h1 −ζ5 u 714h1 −ζ5u 1187b1 ζ5 u 1216k1 ζ5u

4 Table 1: Conductor fχ = 11 with χ(2) = ζ5, ∆Kχ = 11 .

Example 45. The examples are even easier to ﬁnd for cubic characters χ. As before, we will look at curves with

χ χ E(Q) = E(K ) = X = X χ = c (E/Q)= c (E/K )=1. | | | | | E/Q| | E/K | p v p v Y Y All of the curves we look at will satisfy the conditions of Theorem 38, and thus by the same theorem we can predict the L-values up to sign. How to predict the sign is unclear, even for curves with the same conductor. 26 VLADIMIRDOKCHITSER, ROBERT EVANS,HANNEKE WIERSEMA

E L (E,χ) E L (E,χ) E L (E,χ) E L (E,χ) E L (E,χ) 1356d1 ζ3 3264r1 −ζ3 3540a1 −ζ3 4800i1 −ζ3 1356f1 −ζ3 3264s1 ζ3 3540b1 ζ3 4800bj1 −ζ3 4800bm1 ζ3 2 Conductor fχ = 7 with χ(3) = ζ3 , ∆Kχ = 49. 2 222b1 −1 1392c1 −1 4386c1 −1 9024l1 −ζ3 2 222e1 1 1392j1 1 4386m1 1 9024bf1 ζ3 2 Conductor fχ = 13 with χ(2) = ζ3 , ∆Kχ = 169. 702d1 −1 1443a1 1 5616j1 −1 12096bq1 1 19008u1 −1 702i1 1 1443b1 −1 5616o1 1 12096dc1 −1 19008bh1 1 5616p1 1 12096dd1 1 2 Conductor fχ = 19 with χ(2) = ζ3 , ∆Kχ = 381. 714b1 −1 2453a1 1 8138b1 1 12096x1 ζ3 714h1 1 2453c1 −1 8138c1 −1 12096dc1 ζ3 12096dd1 −ζ3 Conductor fχ = 31 with χ(3) = ζ3, ∆Kχ = 961. 2 2 5885a1 −ζ3 11764a1 −ζ3 12096x1 ζ3 15498h1 −ζ3 16590c1 1 2 2 5885d1 ζ3 11764b1 ζ3 12096bb1 −ζ3 15498i1 ζ3 16590n1 −1 2 12096bn1 ζ3 2 12096cz1 −ζ3 Conductor fχ = 37 with χ(2) = ζ3, ∆Kχ = 1369

Table 2: Modiﬁed L-values for varying Dirichlet characters χ of order three.

In these examples, the curves in each block also have discriminants of the same sign as each other and the same number of points over F3. The first condition ensures that they have the same number of real components. The second condition is motivated by p-adic L-functions, where the interpolation formula for L-values is adjusted by an extra term that depends on E(F ) . | p | Example 46. Here we give a list of curves similar to Example 42. We again take the 3 character χ of order five and conductor 31 defined by χ(3) = ζ5 , and consider curves with conductor coprime to 31 with

χ χ 2 E(Q) = E(K ) = X = c (E/Q)= c (E/K ) = 1 and X χ = 11 . | | | | | E/Q| p v | E/K | p v Y Y We know from Remark 43 that the ideal (L (E,χ)) of is either p p or p p , where OQ(ζ5) 1 2 3 4 p1 = (11, 7+ ζ5), p2 = (11, 8+ ζ5), p3 = (11, 6+ ζ5), p4 = (11, 2+ ζ5) are the primes of Q(ζ5) above 11. For the following list of curves list L (E,χ) splits as p1p2: 216b1, 216c1, 291d1, 443c1, 475a1. For the following list of curves L (E,χ) splits as p3p4: 139a1, 140b1, 267b1, 333d1, 378h1, 432g1, 579a1.

References

[1] A. Bartel, Large Selmer groups over number ﬁelds, Math. Proc. Cambridge Philos. Soc. 148 (2010), no. 1, 73–86. [2] W. Bosma, J. Cannon, C. Playoust, The Magma algebra system. I: The user language, J. Symb. Comput. 24, No. 3–4 (1997), 235–265. [3] D. Burns, D. M. Castillo, On reﬁned conjectures of Birch and Swinnerton-Dyer type for Hasse–Weil– Artin L-series, arxiv: 1909.03959. [4] J. Coates, Motivic p-adic L-functions, in: L-functions and arithmetic (Durham, 1989), 141–172, London Math. Soc. Lecture Note Ser., 153, Cambridge Univ. Press, Cambridge, 1991. [5] V. Dokchitser, Root numbers of non-abelian twists of elliptic curves, Proc. London Math. Soc. (3) 91 (2005), 300–324. ON A BSD-TYPE FORMULA FOR L-VALUES OF ARTIN TWISTS OF ELLIPTIC CURVES 27

[6] T. Dokchitser, V. Dokchitser, Regulator constants and the parity conjecture, Invent. Math. 178 (2009), 23–71. [7] P. Deligne, Valeurs de fonctions L et périodes d’intégrales, in: Automorphic forms, representations and L-functions, Part 2 (ed. A. Borel and W. Casselman), Proc. Symp. in Pure Math. 33 (AMS, Providence, RI, 1979) 313–346. [8] J. Martinet, Character Theory and Artin L-functions, in: Algebraic Number Field (ed. A. Frohlich), Proc. of the Durham Symp. (Academic Press, London, 1975). [9] K. Matsuno, Elliptic curves with large Tate–Shafarevich groups over a number field, Math. Research Letters 16 (2009), no. 3, 449–461. [10] D. Rohrlich, The vanishing of certain Rankin-Selberg convolutions, in: Automorphic Forms and Analytic Number Theory, 123–133. Les publications CRM, Montreal (1990). [11] J. H. Silverman, Advanced topics in the arithmetic of elliptic curves, GTM 151, Springer, Berlin (1994). [12] G. Stevens, Stickelberger elements and modular parametrizations of elliptic curves, Invent. Math. 98 (1989), no. 1, 75–106. [13] J. Tate , On the conjectures of Birch and Swinnerton–Dyer and a geometric analog, Séminaire Bourbaki, 18e année, 1965/66, no. 306. [14] J. Tate, Number theoretic background, in: Automorphic forms, representations and L-functions, Part 2 (ed. A. Borel and W. Casselman), Proc. Symp. in Pure Math. 33 (AMS, Providence, RI, 1979) 3–26. [15] O. Venjakob, From the Birch and Swinnerton-Dyer conjecture over the equivariant Tamagawa number conjecture to non-commutative Iwasawa theory - a survey, in: L-functions and Galois representations, London Math. Soc. Lecture Note (Cambridge University Press, 2007) 333–380. [16] H. Wiersema, C. Wuthrich, Integrality of twisted L-values of elliptic curves, arXiv:2004.05492.

University College London, 25 Gordon Street, London WC1H 0AY, UK E-mail address: [email protected]

King’s College London, Strand, London WC2R 2LS, UK E-mail address: [email protected]